🚀 Get Started

📝 Context

This repository contains the code for the work done by Télécom Physique Strasbourg students on the generation of realistic tweets using LLMs (a study committed by Euro Information).

[poster placeholder]

✏️ Setup

Please note we do not officially support Windows or MacOS, but we do provide some instructions for those who want to use it on these platforms.

You do not explicitly need a conda environment for the bot to run. But it is always recommended nontheless, especially because the next LTS of Ubuntu won't let users pip-install anything without a virtual environment. At the time of writing, this app python >= 3.10 to run.

First, install the dependencies :

conda create -n xsg python=3.10  # get python 3.10 for torch
conda activate xsg               # env
pip install -r requirements.txt  # install pip deps
python xs-gen do doctor          # check install

Then, you need to provide a list of topics and sentiments to the app. You can do so by creating two files in the assets/config folder, topics.txt and sentiments.txt. Each file should contain one topic or sentiment per line.

# assets/config/topics.txt
pizza
burger
fries
salad
# assets/config/sentiments.txt
positive
negative
neutral

You can then run the app (more details on the commands next pages) :

python xs-gen.py run

⏱️ Just the commands

conda env create -f environment.yml
conda activate xsg

printf "positive\nnegative\nneutral\n" > assets/config/sentiments.txt
printf "pizza\nMia Khalifa\n" > assets/config/topics.txt

python xs-gen.py run